Delft University of Technology
Assessing Balance Control After Minor Stroke
Moving from Laboratory towards Clinic
Schut, I.M.
DOI
10.4233/uuid:67ba116d-6770-468a-ad5a-5135ac646290
Publication date
2020
Document Version
Final published version
Citation (APA)
Schut, I. M. (2020). Assessing Balance Control After Minor Stroke: Moving from Laboratory towards Clinic.
https://doi.org/10.4233/uuid:67ba116d-6770-468a-ad5a-5135ac646290
Important note
To cite this publication, please use the final published version (if applicable).
Please check the document version above.
Copyright
Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy
Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.
This work is downloaded from Delft University of Technology.
AFTER MINOR STROKE
Moving from Laboratory towards Clinic
Colofon
Assessing Balance Control After Minor Stroke Moving from Laboratory towards Clinic Ingrid Marjolein Schut
ISBN/EAN: 978-94-6375-955-7
Copyright © 2020 Ingrid Marjolein Schut
All rights reserved. No part of this thesis may be reproduced, stored or transmitted in any way or by any means without the prior permission of the author, or when applicable, of the publishers of the scientific papers.
Cover design: Ingrid Marjolein Schut & Birgit Vredenburg Layout by Birgit Vredenburg, persoonlijkproefschrift.nl Printing: Ridderprint | www.ridderprint.nl
This research was funded by the Netherlands Organisation for Health Research and Development (ZonMw)under the research programme Innovative Medical Devices Initiative (IMDI) NeuroControl, nr. 104003014, project Move On: A novel balance-testing device to improve mobility after stroke.
ASSESSING BALANCE CONTROL
AFTER MINOR STROKE
Moving from Laboratory towards Clinic
Dissertation
For the purpose of obtaining the degree of doctor at Delft University of Technology
by the authority of the Rector Magnificus Prof.dr.ir. T.H.J.J. van der Hagen Chair of the Board for Doctorates
To be defended publicly
Wednesday 30 September 2020 at 10:00 am
by
Ingrid Marjolein SCHUT
Master of Science in Biomedical Engineering Born in Delft
Composition of the doctoral committe: Rector Magnificus Chairman
Prof. dr. H. van der Kooij Delft University of Technology, promotor Dr. ir. A.C. Schouten Delft University of Technology, promotor
Dr. V. Weerdesteyn Radboud University Medical Center, copromotor Independent member:
Prof. dr. T.J.M. van der Cammen Delft University of Technology Prof. dr. ir. M. Mulder Delft University of Technology
Dr. C.G.M. Meskers Amsterdam University Medical Center Dr. K. Meijer University of Maastricht
Financial support by the Dutch Heart Foundation for the publication of this thesis is gratefully acknowledged.
Summary 6
Samenvatting 10
Ch. 1 Introduction 17
PART I – Assessment of balance control using system identification 29 Ch. 2 Compliant support surfaces affect sensory reweighting 31
during balance control
Ch. 3 Effect of amplitude and number of repetitions of the 47 perturbation on system identification of human balance
control during stance
Ch. 4 Measuring joint stiffness on a treadmill using system 71 identification: no need for horizontal forces
PART II – Balance assessment in minor strokes 87
Ch. 5 Minor stroke, serious problems: the impact on balance 89 and gait capacity, fall rate, and physical activity
Ch. 6 Detection of balance control asymmetries in people 109 with minor stroke
Ch. 7 General discussion and conclusion 135
APPENDIX 147
Acknowledgements 149
About the author 153
Summary Summary
Summary
In this thesis, we determined how system identification can be integrated in the clinic to assess human balance control. Adequately balance assessment is important in order to detect people with high fall risks and to optimize balance training, which would ultimately result in better rehabilitation and thus less falls. Current clinical tests suffer from ceiling effects, are subjective and do not provide insight in the underlying mechanisms. System identification techniques seem promising, but they depend on large, expensive and complex devices such as motion platforms and motion capture cameras, and are therefore less suitable for clinical use. In part 1, we first focussed on the technical and methodological characteristics of the system identification techniques. In part 2, we used the resulting system identification method, whereby the treadmill applied support surface perturbations, on a minor stroke population to evaluate subtle changes in balance control of the paretic leg.
System identification uses dedicated perturbations to unravel the underlying neurophysiological mechanisms of balance control. By applying support surface rotations, the sensory reweighting of proprioceptive information can be identified. Support surface translations enable the investigation of the stabilizing mechanism. In Chapter 2, system identification was implemented with the bilateral ankle perturbator (BAP), which applies support surface rotations around the ankle joint. A new application of the BAP was used investigate the effect of compliant support surfaces. Compliant surfaces, e.g. foam mats, are currently used in the training as tool for diagnosis and training. With the BAP continuous disturbance torques were applied in 9 trials; three levels of support surface compliance, combined with three levels of desired support surface rotation amplitude. The corrective ankle torques, in response to the rotations, were assessed in frequency response functions. Low frequency magnitude, i.e. the average frequency response function magnitudes in a lower frequency window, represents the sensory reweighting. As found previously, an increase in support surface rotation amplitude leads to a decrease in the lower frequency magnitude, i.e. down weighting of proprioceptive information. We found that this down weighting effect is less when the support surface is more compliant. In other words, the sensory reweighting of proprioceptive information due to support surface rotation amplitude is less on more compliant support surfaces. Therefore, it might be interesting to use foam mats with different compliances, since a wider range of sensory reweighting will be trained.
In Chapter 3 and 4, system identification is combined with a treadmill, which applies support surface translations. The methodological characteristics were studied in
Chapter 3, investigating the effect of the amplitude and number of repetitions of the perturbation signal on the identification of the neuromuscular controller (NMC). Chapter 4 focused on the treadmill characteristics. In chapter 3, healthy participants were asked to stand on a treadmill while small continuous support surface translations were applied in the form of a periodic multisine signal. The perturbation amplitude varied over seven conditions between 0.02 and 0.20 m peak-to-peak (ptp), where 6.5 repetitions of the multisine signal were applied for each amplitude, resulting in a trial length of 130 sec. For one of the conditions, 24 repetitions were recorded. The recorded external perturbation torque, body sway and ankle torque were used to calculate both the relative variability of the frequency response function of the NMC, i.e. a measure for precision, which depends on the noise-to-signal ratio and the nonlinear distortions. In general, the nonlinear contributions were low and, for the ankle torque did not vary with perturbation amplitude. Results showed that the perturbation amplitude should be minimally 0.05 m ptp, but higher perturbation amplitudes are preferred since they result in a higher precision, due to a lower noise-to-signal ratio. There is, however, no need to further increase the perturbation amplitude than 0.14 m ptp. More repetitions improves precision, but the number of repetitions minimally required, eventually depends on the perturbation amplitude and the preferred precision.
Chapter 4 validated the joint stiffness estimation of an inverted pendulum with known stiffness using system identification and support surface translations. Second, the contribution of horizontal ground reaction forces on the estimation was investigated. Ankle torque and resulting frequency response functions, which describes the dynamics of the stabilizing mechanism, were calculated by both including and excluding horizontal ground reaction forces. Results showed that the joint stiffness of an inverted pendulum estimated using system identification is comparable to the joint stiffness estimated by a regression method. Secondly, within the induced body sway angles, the ankle torque and frequency response function of the stabilizing mechanism dynamics calculated by both including and excluding horizontal ground reaction forces are similar. Especially for the lower frequencies, where stiffness dominates the frequency response function, the absolute relative errors are small (<1%). Therefore, the horizontal ground reaction forces can be omitted in calculating the ankle torque and frequency response function of the stabilizing mechanism dynamics. Chapter 3 and 4 show that the treadmill could be used to assess balance control, when minor adjustments are made to treadmills currently available in the clinic; vertical forces, i.e. vertical ground reaction forces and CoP in sagittal plane, should be recorded by a dual force plates, and continuous forward-backward translations of the belt should be possible.
In Part II, balance control of people with a minor stroke was assessed. In chapter 5, we indicated that people after minor stroke have an elevated fall risk, are less physically active, and show persistent balance impairments compared to their healthy counterparts, even though they have (almost) complete clinical recovery of leg motor impairments. Fugl-Meyer Lower extremity assessment and Mini-Balance Evaluation Systems Test (Mini-BESTest) were performed on a total of 64 minor stroke participants and 50 age-matched controls. Fall rates and daily physical activity levels were conducted in a follow-up period of six months. Minor stroke participants fell almost twice as often as controls (1.1 vs. 0.52 falls per person per year). The total time of physical activity was not significantly different for minor stroke participants compared to the controls, whereas the total intensity of physical activity was 6% lower for minor stroke participants. Minor stroke participants also scored significantly lower on the Mini-BESTest (24.2±2.3 vs. 26.1±2.1 points), indicating impaired static and dynamic balance, even when they scored maximal on the Fugl-Meyer Lower extremity assessment. There results indicate that individuals in the chronic phase after minor stroke with (almost) complete clinical recovery of leg motor impairments fall more often and show substantial impairments in dynamic balance control compared to healthy controls, pointing at an important unmet clinical need in this population.
Impaired balance and mobility of minor stroke participants, as shown in Chapter 5, might be caused by subtle changes in balance control of the paretic ankle, indicating an asymmetric balance control strategy. In Chapter 6, we investigated whether subtle control asymmetries can be demonstrated in minor stroke participants, in which balance measure these subtle changes in control asymmetries are most apparent and under which condition they could best be detected. Balance control of 54 minor stroke and 37 control participants was assessed. In static and perturbed conditions, both performed with eyes open and eyes closed, 11 balance measures were conducted: centre of pressure (CoP) related measures, torque related measures and the dynamic balance contribution, i.e. the contribution of the paretic leg to the total frequency response function, which describes the dynamics of the stabilizing mechanism in terms of a magnitude and phase. To investigate whether changes in control asymmetry are more apparent in a combination of the balance measure symmetry indices, principal component analysis was performed for each condition. Our results show that the individual balance measure symmetry indices were not significantly different between minor stroke and control participants. However, when the symmetry indices, measured during the perturbed condition with eyes closed, were combined according to principal component analysis, the resulting first principal component score was significantly higher for stroke participants. The first principal component includes symmetry indices of the mean dynamic balance contribution and root-mean-square of anterior-posterior CoP position and velocity,
i.e. symmetry indices indicating control in anterior-posterior direction. Replacing the dynamic balance contribution by the root-mean-square of the torque resulted in similar outcomes. This chapter indicates that subtle changes in control asymmetry of minor stroke participants cannot be demonstrated by measuring individual balance measures, but can be demonstrated by combining the dynamic balance contribution, CoP position and CoP velocity when measured during the perturbed trial with eyes closed.
In Chapter 7 we discuss the key findings and clinical implementation of this thesis, and elaborate on the considerations and future recommendations. Throughout this thesis we showed how the BAP and the treadmill can be used in combination with system identification to assess balance control, i.e. sensory reweighting and the stabilizing mechanism. This indicates that the large, expensive and complex motion platforms can be replaced, which makes it possible to assess balance control in the clinic. The advantage of treadmills is that they are already used in the clinic for training and can easily be adjusted such that our system identification method can be applied. First, the treadmill should be able to apply continuous forward-backward translations of the support surface with amplitudes in a range of 0.05-0.14 m ptp. This might require a stronger motor, if not already present, and small adjustments to the software. Second, vertical ground reaction forces should be recorded of both feet separately. This requires the replacement of a single force plate with dual force plates. We showed that using this setup, subtle changes in balance control of the paretic ankle of minor stroke patients can be demonstrated by combining the dynamic balance contribution, posterior CoP position and anterior-posterior CoP velocity, when measured during perturbed conditions with eyes closed. Future research should point out whether these methods could ultimately be used for the detection of people with high fall risks and assessment of training effects.
Samenvatting Samenvatting
Samenvatting
In dit proefschrift hebben we vastgesteld hoe systeem identificatie geïntegreerd kan worden in de kliniek om menselijke balans control te beoordelen. Een goede balansbeoordeling is belangrijk om mensen met een hoog valrisico te detecteren en om balanstraining te optimaliseren, wat uiteindelijk zou leiden tot betere revalidatie en dus minder valpartijen. Huidige klinische testen hebben last van plafondeffecten, zijn subjectief en bieden geen inzicht in de onderliggende mechanismen. Systeemidentificatie lijkt veelbelovend, maar is afhankelijk van grote, dure en complexe apparaten zoals bewegingsplatformen en motion-capture camera’s en is daarom minder geschikt voor klinisch gebruik. In deel 1 van dit proefschrift hebben we ons gericht op de technische en methodologische kenmerken van systeemidentificatie technieken. In deel 2 hebben we de resulterende systeemidentificatie methode, waarbij de loopband zorgde voor verstoringen van de ondergrond, toegepast bij mensen met een lichte beroerte om subtiele veranderingen in balanscontrole van het paretische been te evalueren.
Systeemidentificatie maakt gebruik van specifieke verstoringen om de onderliggende neurofysiologische mechanismen van balans controle te ontrafelen. Door rotaties van de ondergrond toe te passen, kan de sensorische herweging van proprioceptieve informatie worden geïdentificeerd. Translaties van de ondergrond maken het mogelijk om het stabilisatiemechanisme te onderzoeken. In hoofdstuk 2 werd systeemidentificatie geïmplementeerd met de ‘bilateral ankle perturbator’ (BAP), die rotaties van de ondergrond rond het enkelgewricht toepast. Een nieuwe toepassing van de BAP werd gebruikt om het effect van de ondergrond compliantie te onderzoeken. Flexibele ondergronden, b.v. balansmatten, worden vandaag de dag gebruikt als hulpmiddel voor diagnose en training. Met de BAP werden continue verstoringsmomenten toegepast in negen trials; drie ondergronden met een verschillende compliantie, gecombineerd met drie gewenste ondergrond amplitudes. De corrigerende enkelmomenten, als reactie op de rotaties, werden beoordeeld in frequentie responsiefuncties. De laagfrequente magnitude, d.w.z. de gemiddelde frequentie responsfunctie magnitudes in een laag frequentievenster, geeft de sensorische herweging aan. In overeenkomst met de literatuur, leidt een toename van de rotatie-amplitude van de ondergrond tot een afname van de laagfrequente, d.w.z. een lagere weging van proprioceptieve informatie. We hebben geconstateerd dat dit neerwaartse wegingseffect minder is naarmate de ondergrond een hogere compliantie heeft. Met andere woorden, de sensorische herweging van proprioceptieve informatie door de rotatie-amplitude van de ondergrond, is minder op een zachtere ondergrond. Het kan daarom interessant zijn om schuimmatten met verschillende stijfheden te gebruiken, aangezien een groter bereik van sensorische herweging wordt getraind.
In hoofdstuk 3 en 4 wordt systeemidentificatie gecombineerd met een loopband die translaties van de ondergrond toepast. De methodologische kenmerken werden bestudeerd in hoofdstuk 3, waarbij het effect van de amplitude en het aantal herhalingen van het verstoringssignaal op de identificatie van de neuromusculaire controller (NMC) werd onderzocht. Hoofdstuk 4 richtte zich op de karakteristieken van de loopband. In hoofdstuk 3 werd aan gezonde deelnemers gevraagd op een loopband te staan, terwijl kleine continue translaties van de ondergrond werden toegepast in de vorm van een periodiek multisinus signaal. De verstoringsamplitude varieerde over zeven condities tussen 0,02 en 0,20 m piek-tot-piek (ptp), waarbij 6,5 herhalingen van het multisinus signaal werden toegepast voor elke amplitude, resulterend in een trial lengte van 130 sec. Voor een van de condities werden 24 herhalingen opgenomen. Het gemeten externe moment, body sway, en enkel moment werden gebruikt om zowel de relatieve variabiliteit van de frequentieresponsfunctie van de NMC, wat een maat voor precisie is en die afhankelijk is van de ruis-signaalverhouding, als de niet-lineaire contributies te berekenen. Over het algemeen waren de niet-lineaire bijdragen laag en varieerde, voor het enkelmoment, niet met verstoringsamplitude. De resultaten toonden aan dat de verstoringsamplitude minimaal 0,05 m ptp zou moeten zijn, maar hogere verstoringsamplitudes hebben de voorkeur omdat ze resulteren in een hogere precisie vanwege een lagere ruis-signaalverhouding. Het is echter niet nodig om verstoringsamplitudes hoger dan 0,14 m ptp te gebruiken. Meer herhalingen verbeteren de precisie, maar het minimaal vereiste aantal herhalingen hangt uiteindelijk af van de verstoringsamplitude en de gewenste precisie.
Hoofdstuk 4 valideert de stijfheidsschatting van een omgekeerde slinger met bekende stijfheid, met behulp van systeemidentificatie en translaties van de ondergrond. Ten tweede werd de bijdrage van horizontale grondreactiekrachten aan de schatting onderzocht. Het enkelmoment en de daaruit voortvloeiende frequentie responsiefuncties, die de dynamiek van het stabilisatiemechanisme beschrijven, werden berekend door de horizontale grondreactiekrachten zowel binnen als buiten beschouwing te laten. De resultaten toonden aan dat de stijfheid van een omgekeerde slinger geschat met behulp van systeemidentificatie vergelijkbaar is met de stijfheid geschat met een regressiemethode. Ten tweede zijn, binnen de geïnduceerde body sway, het enkelmoment en de frequentie responsiefunctie van het stabilisatiemechanisme, berekend door de horizontale grondreactiekrachten zowel binnen als buiten beschouwing te laten, vergelijkbaar. Vooral voor de lagere frequenties, waar stijfheid de frequentie responsiefunctie domineert, zijn de absolute relatieve fouten klein (<1%). Daarom kunnen de horizontale grondreactiekrachten worden weggelaten bij de berekening van het enkelmoment en de frequentie responsiefunctie van de stabilisatiemechanisme
dynamiek. Hoofdstuk 3 en 4 laten zien dat de loopband kan worden gebruikt om balans te meten wanneer kleine aanpassingen worden gedaan aan loopbanden die momenteel in de kliniek beschikbaar zijn; verticale krachten, d.w.z. verticale grond reactie krachten en CoP in het sagittale vlak, moeten worden gemeten door een dubbele krachtplaat, en continue voorwaartse-achterwaartse translaties van de belt moeten mogelijk zijn. In deel II werd de balanscontrole van mensen met een lichte beroerte beoordeeld. In hoofdstuk 5 hebben we aangetoond dat mensen na een lichte beroerte een verhoogd valrisico hebben, fysiek minder actief zijn en aanhoudende balansstoornissen vertonen in vergelijking met hun leeftijdsgenoten, ook al zijn ze (bijna) volledig klinisch hersteld van motorische stoornissen in het been. De ‘Fugl-Meyer Lower extremity assessment’ en ‘Mini-Balance Evaluation Systems Test’ (Mini-BEST) werden uitgevoerd op een 64 deelnemers die een lichte beroerte hebben gehad en 50 leeftijdsgebonden controles. Valcijfers en dagelijkse fysieke activiteit niveaus werden in een follow-up periode van zes maanden bijgehouden. Deelnemers met een lichte beroerte vielen bijna twee keer zo vaak als gezonde controles (1,1 vs. 0,52 valpartijen per persoon per jaar). De totale tijd van lichamelijke activiteit was niet significant verschillend voor deelnemers met een lichte beroerte in vergelijking met de controles, terwijl de totale intensiteit van de lichamelijke activiteit 6% lager was voor deelnemers met een lichte beroerte. Deelnemers met een lichte beroerte scoorde ook significant lager op de Mini-BEST (24,2 ± 2,3 vs. 26,1 ± 2,1 punten), wat duidt op verminderde statische en dynamische balans, zelfs wanneer ze maximaal scoorden op de ‘Fugl-Meyer Lower extremity assessment’. De resultaten geven aan dat personen in de chronische fase na een lichte beroerte met (bijna) volledig klinisch herstel van motorische stoornissen aan het been vaker vallen en aanzienlijke stoornissen in dynamische balanscontrole vertonen in vergelijking met controles, wat wijst op een belangrijke onvervulde klinische behoefte in deze populatie.
Een verminderde balans en mobiliteit van deelnemers met een lichte beroerte, zoals weergegeven in hoofdstuk 5, wordt wellicht veroorzaakt door subtiele veranderingen in de balanscontrole van de paretische enkel, wat duidt op een asymmetrische balanscontrolestrategie. In hoofdstuk 6 hebben we onderzocht of subtiele control asymmetrie kan worden gedetecteerd bij deelnemers met ene lichte beroerte, in welke mate deze subtiele veranderingen in de control asymmetrie het meest zichtbaar zijn en in welke conditie ze het best kunnen worden gedetecteerd. Balanscontrole van 54 deelnemers met een lichte beroerte en 37 leeftijdsgebonden controles werd beoordeeld. In statische en verstoorde condities, beide uitgevoerd met open en gesloten ogen, werden 11 balansmaten gemeten: centre of pressure (CoP) gerelateerde maten, moment gerelateerde maten en de dynamic balance contribution, d.w.z. de bijdrage
van het paretische been aan de totale frequentie responsiefunctie, die de dynamiek van het stabilisatiemechanisme beschrijft in termen van een magnitude en fase. Om te onderzoeken of veranderingen in de control asymmetrie duidelijker naar voren komen in een combinatie van de balansmaat symmetrie-indices, werd voor elke conditie een principal component analyse uitgevoerd. Onze resultaten tonen aan dat de individuele balansmaat symmetrie-indices niet significant verschilden tussen deelnemers met een lichte beroerte en controledeelnemers. Echter, wanneer de symmetrie-indices werden gemeten tijdens de verstoorde conditie met gesloten ogen, en werden gecombineerd volgens de principal component analyse, was de resulterende score van de eerste principal component significant hoger voor de deelnemers met een lichte beroerte in vergelijking met de controles. De eerste principal component bestond uit symmetrie-indices van de gemiddelde dynamic balance contribution en de root-mean-square van de anterior-posterior CoP-positie en snelheid, d.w.z. symmetrie-indices die de controle in anterior-posterior richting aangeven. Het vervangen van de dynamic balance contribution door de root-mean-square van het enkel moment resulteerde in gelijke resultaten. Dit hoofdstuk geeft aan dat subtiele veranderingen in de control asymmetrie van mensen met een lichte beroerte niet aangetoond kunnen worden door het meten van individuele balansmaten, maar wel aangetoond kunnen worden wanneer de dynamic balance contribution, CoP positie en CoP snelheid worden gemeten tijdens de verstoorde trial met gesloten ogen.
In hoofdstuk 7 bespreken we de belangrijkste bevindingen en de klinische implementatie van dit proefschrift, en gaan we dieper in op de overwegingen en toekomstige aanbevelingen. In dit proefschrift hebben we laten zien hoe de BAP en de loopband kunnen worden gebruikt in combinatie met systeemidentificatie om de balanscontrole, d.w.z. de sensorische herweging en het stabilisatiemechanisme, te evalueren. Hiermee geven we aan dat de grote, dure en complexe bewegingsplatformen vervangen kunnen worden, wat het mogelijk maakt om de balanscontrole in de kliniek te beoordelen. Het voordeel van loopbanden is dat ze al in de kliniek gebruikt worden voor training en eenvoudig kunnen worden aangepast zodat onze systeemidentificatiemethode kan worden toegepast. Ten eerste moet de loopband in staat zijn om continue voorwaartse en achterwaartse translaties van de belt toe te passen met amplitudes in een bereik van 0,05-0,14 m ptp. Dit kan een sterkere motor vereisen, als deze nog niet aanwezig is, en kleine aanpassingen aan de software. Ten tweede moeten de verticale grondreactiekrachten van beide voeten afzonderlijk gemeten kunnen worden. Dit vereist de vervanging van een enkele krachtplaat door een dubbele krachtplaat. We toonden aan dat met behulp van deze opstelling, subtiele veranderingen in de balans controle van de paretische enkel van mensen met een lichte beroerte aangetoond kunnen worden door het combineren
Samenvatting Samenvatting
van de dynami balance contribution, anterior-posterior CoP positie en anterior-posterior CoP velocity als deze gemeten worden tijdens de verstoorde conditie met gesloten ogen. Toekomstig onderzoek moet uitmaken of deze methoden uiteindelijk ook gebruikt kunnen worden om mensen met een hoog valrisico te detecteren en de trainingseffecten te beoordelen.
1
Introduction Chapter 1
Rationale
Balance control (Box 1) can be impaired due to aging or pathologies such as Parkinson’s disease or stroke (Box 2), which increases the risk of falls [1,2,3]. The assessment of balance control in the clinical setting is important to detect people with high fall risk. If these people are adequately detected, proper rehabilitation can be provided. In addition, balance assessment is important to investigate the individual effect of balance training. The training can be adjusted to the individual needs, which is expected to ultimately result in better rehabilitation and thus less falls.
Currently, balance assessment in clinical settings consists of performance-based tests such as the Berg Balance Scale [15] and the Mini-BEST [16], which involve a variety of tasks, including transferring from sit-to-stance, standing still on one or two legs, normal walking and turning. These tests suffer from important limitations. First, some frequently used clinical tests, e.g. the Berg Balance Scale, have substantial ceiling effects [17], and are hence insensitive to more subtle impairments that still have an impact on functioning in more advanced activities of daily life. Second, the assessments are subjective, as the execution can differ from test to test and doctor to doctor or results might be influenced by the patient’s fear of falling, leading to unclear results. Third, they do not provide insight into the neurophysiological mechanisms underlying impaired performance [4,18]. These mechanisms may differ greatly between individual patients and, consequently, warrant different treatment approaches. Therefore, there is a need for objective tests that do not suffer from ceiling effects and are capable of providing insight into the underlying mechanisms.
An alternative method to assess balance control is posturography. Posturography assesses the excursions of the center of pressure (CoP) which reflect the amount of body sway, as well as the generated corrective torques to keep the center of mass (CoM) above the base of support [19]. This makes it possible to detect balance or gait abnormalities in an objective assessment. In addition, impairments in the sensory system can be assessed by manipulation of the proprioceptive information, i.e. standing on a firm or compliant support surface, or by manipulation of vision, i.e. conditions with eyes open and eyes closed [18]. Although posturography is objective and has no ceiling effects, it still lacks the ability to distinguish between the underlying neurophysiological mechanisms due to the closed loop and compensation strategies used.
Box 1: Balance Control System
Human balance, i.e. keeping the body in an upright position, depends on the appropriate functioning of the balance control system, which comprises many subsystems, such as the sensory, nervous and motor systems, interacting with each other in a closed loop. In a simplified model of balance control the body, represented by an inverted pendulum, is constantly challenged by disturbances such as pushes (torque perturbations), or sensory noise (sensory perturbations) [4]. Information about changes in body position (body sway θ), is detected by the neuromuscular controller: the information is weighted in relation to its reliability by sensory systems, i.e. sensory reweighting, and then integrated by the neural controller which sends signals to the motor system, creating torques (T). The torques result in a change in body position, which is again detected by the sensory system. Depending on the disturbance, the balance control system is used in different strategies. When the perturbations are small, the motor system creates a corrective ankle torque, sufficient to keep the body in an upright position. This is known as the ankle strategy. When disturbances are larger, the created corrective ankle
torque is not sufficient to keep the body in an upright position, an additional corrective hip torque is created, i.e. the hip strategy, or a corrective step is required, i.e. stepping strategy [5]. Another strategy is to counteract the disturbance by arm movements [6]. In this thesis we focus on the ankle strategy.
Recently, researchers have developed promising experimental tests that are objective, have no ceiling effects and are capable of revealing the mechanisms underlying impaired standing balance. System identification (Box 3) was used in combination with dedicated disturbance signals, which actively perturbed the body while standing [4,18,20-24]. In some studies using system identification, participants’ stance was perturbed by small continuous random rotations of the ankle. The rotations, manipulating the proprioceptive information, are often combined with manipulations of the visual information, i.e. trials are repeated with eyes open and eyes closed [18,21]. By perturbing the sensory systems and measuring the sensitivity functions, i.e. the response of the body in terms of body sway and ankle torque to the perturbation, sensory reweighting can be identified.
Box 2: Stroke
In the Netherlands, an estimated 41,000 people are affected by a stroke each year. A stroke is caused by a reduction of blood flow to the brain. There are two types of strokes: ischemic and hemorrhagic. The most common type is ischemic, occurring in 87% of the cases, and is caused by a blockage of the blood vessel. Hemorrhagic strokes, occurring in 13% of the cases, are caused by a ruptured blood vessel, causing intracranial swelling and pressure. Both
types of stroke lead to the damaging or death of brain tissue. Due to combined sensori-motor and cognitive deficits such as paresis, sensory loss, defective coordination and perceptual and attentional problems, a stroke can lead to impaired balance control during standing and walking [7,8]. The impaired balance of stroke survivors not only has a great impact on independent mobility [9] but is also the most important risk factor for falls [3, 10,11]. Throughout the post-stroke life span, fall risk for people after stroke is three to ten times greater than for community-dwelling people of the same age [3]. The consequences of these falls are also more severe. For instance, individuals with stroke are three times more likely to sustain a hip fracture due to a fall than people without previous stroke, and they lose independent mobility or even die after a hip fracture more often [3,12]. Depending on the severity, a stroke can results in death, loss of specific abilities such as memory or muscle control, or best case scenario in no deficits at all. Almost half of the patients surviving a stroke have only minor or no obvious physical impairments during the acute phase (<6 months), and are therefore discharged from the hospital without receiving inpatient rehabilitation; subtle changes in their balance control were not evident [13]. Yet, preliminary findings show that these people may still have balance impairments during the chronic phase (>6 months), and are therefore more likely to fall [7, 14].
In other studies using system identification, perturbations were applied by small continuous random anterior-posterior translations in the range of 6-16 cm, delivered by a moveable platform on which the participants had to stand [20, 22]. The amplitude of the platform was based on the balance capacity of the participants; it was large enough to detect a response and small enough to maintain standing. By analyzing the corrective torques and sway as a response to the perturbations, the control of the central nervous system was identified [4,22]. Since previous research showed that system identification enables the estimation of neuromuscular control deteriorations in stroke and Parkinson’s
disease [1,22], it is thought that system identification will be a powerful tool to detect the underlying balance control impairments of stroke patients, including those with minor residual impairments. Although system identification seems promising, there is one major drawback; it depends on large, expensive and complex devices such as motion platforms (Figure 2) and motion capture cameras, and are therefore not suitable for clinical use.
Figure 2: Motion platform used for system identification methods to assess balance control. Source: Journal of Neurophysiology - doi:10.1152/jn.01008.2011
Source: enableme.org.au
Introduction Chapter 1
Box 3: System identification to assess balance Control
Underlying mechanisms of balance control interact in a closed-loop, which makes it impossible to determine cause and effect. System identification is a method to open this loop by identifying the dynamics of each underlying system, thereby disentangling cause and effect. This requires dedicated perturbation signals such as sensory perturbations, e.g. perturbing proprioceptive information or eliminating vision, or force perturbations, e.g. pushes against the body. As humans respond differently to either fast (high frequency) or slow (low frequency) disturbances, the perturbation signals are often applied as unpredictable periodic multisines, including a wide range of frequencies. By transforming the measured signals (i.e. joint angles and joint torques) to the frequency domain using Fourier transformation, the system can be described by a frequency response function (FRF). The FRF described the dynamics in terms of a magnitude (amount) and phase (timing) of the response to the disturbances as a function of frequency.
Recently, the Bipedal Ankle Perturbator (BAP) has been developed to apply support surface rotations [2,24]. The BAP consists of two force plates that induce rotations around the ankle. As for the support surface translations, a possible solution would be to perform system identification in combination with a treadmill. Both the BAP and the treadmill are small, cheap and easy to use, thereby simplifying the integration of the laboratory into the clinic. This would make it possible to perform objective test unravelling the underlying neurophysiological mechanisms of impaired balance control in the clinic. However, the application of the BAP to apply specific compliances of the support surface to measure sensory reweighting is not yet tested. In addition, it is not yet known how to use system identification in combination with a treadmill. Furthermore, it is not yet known whether balance measures derived from system identification are more sensitive than
measures derived from posturography or clinical tests. As balance impairments in minor stroke patients could be underestimated due to compensation of the non-paretic leg, it is important to investigate what balance measure could best be used to evaluate subtle balance impairments in minor stroke patients and under which condition they should be measured.
Problem statement and aim
It is important to adequately assess balance in the clinic to detect people with high fall risks and to investigate the individual effect of balance training, which may ultimately result in better rehabilitation and fewer falls. For this purpose, system identification techniques seem promising, but they depend on large, expensive and complex devices such as motion platforms and motion capture cameras, and are therefore not yet suitable for clinical use.
Accordingly, the aim of this thesis, which is part of the Move On project (Box 4), is to determine how system identification can be integrated in the clinic to assess balance control of minor strokes. First we investigate how we can use system identification to assess balance control, thereby opening the closed-loop by applying perturbations. Second, we apply the results to evaluate subtle changes in the stabilizing mechanism of the paretic leg of people after minor stroke.
Outline
In Part I, we investigated how system identification can be used to assess human balance control. Chapter 2 focuses on the assessment of the contribution of sensory systems to balance control. We use support surface rotations applied by a bilateral ankle perturbator (BAP) to study the effect of compliant support surfaces such as foam mats to sensory reweighting. Chapter 3 and 4 investigate the methodology when using a treadmill in combination with system identification. In Chapter 3, focusing on the perturbation signal characteristics, we investigate the effect of amplitude and the number of repetitions of the perturbation signal on the precision and nonlinear distortions of the frequency response function, thereby providing guidance when choosing appropriate perturbation settings.
Box 4: Move On Project
The work of this thesis was part of the Move On project, a project within the IMDI NeuroControl consortium, which was funded by The Netherlands Organization for Health Research and Development (grant number: 104003014). Move On was a collaboration among Delft University of Technology and Radboud University Medical Center working in close collaboration with industrial partners (Motek Medical, 2MEngineering) and clinical partners (Sint Maartenskliniek, Tolbrug Specialistische Revalidatie, Pieter van Foreest, Amstelland Fysiotherapie). In a cohort of people after a minor stroke, the aim was to investigate whether balance tests assessing standing balance, stepping balance and step adjustment capacity, are sensitive enough to detect subtle balance impairments. During the project, several balance tests were implemented on the N-Mill. The N-Mill is a
treadmill instrumented with a servomotor, two embedded force plates, and a single belt. Belt movements enable both large discreet perturbations as small continuous perturbations.
Chapter 4, focusing on the treadmill features, validates the use of this device to measure human balance by identifying the dynamics of an single inverted pendulum with fixed characteristics, representing the human body. In addition, this chapter investigates the contribution of horizontal ground reaction forces and debates whether they can be neglected.
In Part II we use the results of Part I to assess subtle changes in balance control after minor stroke. First (Chapter 5), we emphasize that for minor strokes, although discharged from the hospital with few or no obvious impairments, balance performance and gait capacity, physical activity and fall rates are affected. Second (Chapter 6), we use the setup resulting from Chapter 3 and 4 to objectively determine in which balance measure the subtle changes in balance control of the paretic leg in people after stroke are most
apparent and under which condition they can best be measured in the clinic, by assessing static balance and balance perturbed by translations of a treadmill belt.
Chapter 7 discusses the key findings and clinical implementation of this thesis, and elaborates on the considerations and future recommendations.
Figure 1: N-Mill. Treadmill designed for the Move On project
Introduction Chapter 1
References
[1] T.A. Boonstra, A.C. Schouten, J.P.P. van Vugt, B.R. Bloem, H. van der Kooij. Parkinson’s disease patients compensate for balance control asymmetry. Journal of Neurophysiology. 2014; 112:3227-3239.
[2] J.H. Pasma, D. Engelhart, A.B. Maier, A.C. Schouten, H. van der Kooij. Changes in sensory reweighting of proprioceptive information during standing balance with age and disease. Journal of Neurophysiology. 2015; 114:3320-3233.
[3] V. Weerdesteyn, M. de Niet, H.J. van Duijnhoven, A.C. Geurts. Falls in individuals with stroke. Journal of Rehabilitation Research and Development. 2008; 45(8):1195-213. [4] D. Engelhart, J.H. Pasma, A.C. Schouten, C.G. Meskers, A.B. Maier, T. Mergner. Impaired
standing balance in elderly: a new engineering method helps to unravel causes and effects. Journal of the American Medical Directors Association. 2014; 15(3):227 e1-6.
[5] B.E. Maki, W.E. McIlroy, G.R. Fernie. Change-in-support reactions for balance recovery. IEEE Engineering in Medicine and Biology Magazine. 2003; 22(2):20-26.
[6] E.H. van Asseldonk, M.G. Carpenter, F.C. van der Helm, H. van der Kooij. Use of induced acceleration to quantify the (de)stabilization effect of external and internal forces on postural responses. IEEE Transactions on Biomedical Engineering. 2007; 54(12):2284-2295. [7] F.A. Batchelor, S.B. Williams, T. Wijeratne, C.M. Said, S. Petty. Balance and gait impairment
in transient ischemic attack and minor stroke. Journal of Stroke Cerebrovasc Diseases. 2015; 24(10):2291-7.
[8] S. Li, G.E. Francisco, P. Zhou. Post-stroke hemiplegic gait: New perspective and insights. Frontiers in Physiology. 2018; 9(1021).
[9] A.A. Schmid, M. van Puymbroeck, P.A. Altenburger, T.A. Dierks, K.K. Miller, T.M. Damush. Balance and balance self-efficacy are associated with activity and participation after stroke: a cross-sectional study in people with chronic stroke. Archives of Physical Medicine and Rehabilitation. 2012; 93(6):1101-1107.
[10] G.B. Campbell, J.T. Matthews. An integrative review of factors associated with falls during post-stroke rehabilitation. Journal of Nursing Scholarship. 2010; 42(4): 395-404.
[11] A. Mansfield, C. Danells, E. Inness, G Mochizuki, W. McIlroy. Between-limb synchronization for control of standing balance in individuals with stroke. Clinical Biomechanics. 2011; 26312-317.
[12] A. Ramnemark, M. Nilsson, B. Borssen, Y. Gustafson. Stroke, a major and increasing risk factor for femoral neck fracture. Stroke. 2000; 31(7):1572-1577.
[13] Nederlands Huisartsen Genootschap. https://www.nhg.org/standaarden/volledig/nhg-standaard-beroerte?tmp-no-mobile=1#idp47104.
[14] J.M.B. Roelofs, K. van Heugten, D. de Kam, V. Weerdesteyn, A.C.H. Geurts. Relationships between affected-leg motor impairment, postural asymmetry, and impaired body sway control after unilateral supratentorial stroke. Neurorehabilitation and Neural Repair. 2018; 32(11): 953-960.
[15] Berg KO, Wood-Dauphinee SL, Williams JI, Maki B. Measuring balance in the elderly: validation of an instrument. Canadian Journal of Public Health. 1992; 83 Suppl 2:S7-11. [16] F. Franchignoni, F. Horak, M. Godi, A. Nardone, A. Giodano. Using psychometric techniques
to improve the Balance Evaluation Systems Test: the miniBESTest. Joural of Rehabilitation Medicine. 2010; 42(4):323-331
[17] L. Blum, N. Korner-Bitensky. Usefulness of the Berg Balance Scale in stroke rehabilitation: a systematic review. Physical Therapy. 2008; 88(5):559-566.
[18] J.H. Pasma, D. Engelhart, A.C. Schouten, H. van der Kooij, A.B. Maier, C.G. Meskers. Impaired standing balance: the clinical need for closing the loop. Neuroscience. 2014; 267:157-165.
[19] D.A. Winter, A.E. Patla, M. Ishac, W.H. Gage. Motor mechanisms of balance during quiet standing. Journal of Electromyography and Kinesiology. 2003; 13(1):49-56.
[20] T.A. Boonstra, A.C. Schouten, H. van der Kooij. Identification of the contribution of the ankle and hip joints to multi-segmental balance control. Journal of Neuroengineering Rehabilitation. 2013; 10:23.
[21] R.J. Peterka. Sensorimotor integration in human postural control. Journal of Neurophysiology. 2002; 88(3):1097-118.
[22] E.H. van Asseldonk, J.H. Buurke, B.R. Bloem, G.J. Renzenbrink, A.V. Nene, F.C. van der Helm. Disentangling the contribution of the paretic and non-paretic ankle to balance control in stroke patients. Experimental Neurolgy. 2006; 201(2):441-51.
[23] H. van der Kooij, E. van Asseldonk, F.C. van der Helm. Comparison of different methods to identify and quantify balance control. Journal of Neuroscience Methods. 2005; 145(1-2):175-203.
[24] A.C. Schouten, T.A. Boonstra, F. Nieuwenhuis, S.F. Campfens, H. van der Kooij. A bilateral ankle manipulator to investigate human balance control. IEEE Transactions on Neural Systems and Rehabilitation Engineering. 2011; 19(6):660-9.
PART
I
Assessment of
balance control
using system
identification
2
Compliant support
surfaces affect sensory
reweighting during
balance control
I.M. Schut*, D.Engelhart*, J.H.Pasma,
R.G.K.M. Aarts, A.C. Schouten
Gait & Posture, 2017, 53: 241-247
*These authors contributed equally to this
Abstract
To maintain upright posture and prevent falling, balance control involves the complex interaction between nervous, muscular and sensory systems, such as sensory reweighting. When balance is impaired, compliant foam mats are used in training methods to improve balance control. However, the effect of the compliance of these foam mats on sensory reweighting remains unclear. In this study, eleven healthy subjects maintained standing balance with their eyes open while continuous support surface (SS) rotations disturbed the proprioception of the ankles. Multisine disturbance torques were applied in nine trials; three levels of SS compliance, combined with three levels of desired SS rotation amplitude. Two trials were repeated with eyes closed. The corrective ankle torques, in response to the SS rotations, were assessed in frequency response functions. Lower frequency magnitudes (LFM) were calculated by averaging the frequency response function magnitudes in a lower frequency window, representative for sensory reweighting. Results showed that increasing the SS rotation amplitude leads to a decrease in LFM. In addition there was an interaction effect; the decrease in LFM by increasing the SS rotation amplitude was less when the SS was more compliant. Trials with eyes closed had a larger LFM compared to trials with eyes open. We can conclude that when balance control is trained using foam mats, two different effects should be kept in mind. An increase in SS compliance has a known effect causing larger SS rotations and therefore greater down weighting of proprioceptive information. However, SS compliance itself influences the sensitivity of sensory reweighting to changes in SS rotation amplitude with relatively less reweighting occurring on more compliant surfaces as SS amplitude changes.
Introduction
Human balance control during stance is continuously challenged by the gravitational field. To maintain an upright posture and prevent falling, balance control involves the complex interaction of nervous, muscular and sensory systems. The central nervous system (CNS) receives feedback about the body orientation from three main sensory systems: the visual, proprioceptive and vestibular system. For each sensory system, the feedback is compared to its reference. The CNS integrates this information and generates an ‘error’, representing deviations of body orientation from upright stance. The error signal of each sensory system is weighted in relation to its reliability; the CNS prefers reliable over less reliable sensory information within an adaptive weighting process termed sensory reweighting [1-3]. Subsequently, the neural controller (NC) generates with a time delay, a corrective, stabilizing torque by selective activation of muscles. This stabilizing torque (together with a torque caused by the intrinsic dynamics of the muscle properties) keeps the body in upright position.
In elderly and in people with neurological, sensory or orthopedic disorders, balance control might be impaired, leading to postural instability and falls [4,5]. People with impaired balance control often undergo functional balance training that is specifically oriented to improve steadiness while standing on compliant surfaces like foam mats [6,7]. It is assumed that sensory reweighting will be trained using these foam mats, since proprioceptive information is disturbed by the compliant support surface [8]. However, the effect of the compliance of these foam mats on sensory reweighting and balance control remains unclear due to a causality problem. The compliant support surface, i.e. the surface in contact with the feet, might have an effect on sensory reweighting induced by the compliance itself, but on the other hand also might have an effect on sensory reweighting provoked by support surface rotations induced by the compliance. System identification techniques in combination with specifically designed external disturbances provide a way to disentangle cause and effect in balance control. By externally exciting the system with an unique input that is not related to the internal signals of the system, a causal relation between the external disturbances and output signals can be created. This ‘opens’ the closed loop and generates informative data about a dynamic system such as balance control [9].
In this paper we investigated the effect of compliant support surfaces, comparable to foam mats, on sensory reweighting of proprioceptive information in balance control using system identification techniques, independent of the effect caused by the change in support surface rotation amplitude. Previous studies showed that increasing the
Compliant support surfaces affect sensory reweighting Chapter 2
amplitude of support surface rotations result in a decrease of the proprioceptive weight (i.e. down weighting), since the proprioception becomes less reliable [1,3,10]. Therefore, we hypothesize that increasing the amplitude of the support surface rotations lowers the reliability of the proprioceptive information and thus results in down weighting of proprioceptive information [11]. Due to the compliance effect of the support surface, an increase in compliance might affect this sensory weighting. If a compliance effect is present, sensory reweighting due to increasing support surface rotation will change for different levels of compliance.
Methods
Subjects
Eleven healthy young volunteers (age: 20-30 years, 8 women, weight: 67.4 ± 8.2 kg, height: 1.85 ± 0.09 m), without any history of balance disorders, musculoskeletal injuries or neurological disorders, participated in this study and gave written informed consent prior to participation. The study was performed according to the principles of the Declaration of Helsinki and approved by the Medical Ethics Committee of Medisch Spectrum Twente, Enschede, the Netherlands.
Apparatus
The Bilateral Ankle Perturbator (BAP, Forcelink B.V., Culemborg, The Netherlands)(Figure 1) consists of two pedals with similar SS rotations around the ankle axis [10]. The BAP was used to mimic compliant support surfaces and to evoke sensory reweighting by rotations of the support surface (SS). The control scheme (Figure 1) shows the control of each compliant support surface of the BAP and the implementation in the human balance control scheme as a torque controlled device; the input is a disturbance torque resulting in a disturbance rotation amplitude as output. For each leg, the reference torque (Tr,L/R) comprises a disturbance torque (Td,L/R) and an exerted ankle torque by the human (Th,L/R) and is translated via the BAP dynamics to a SS rotation amplitude (θSS,L/R). The BAP dynamics consist of a virtual inertia (IV), a virtual damping (BV) and an adjustable virtual stiffness, from here on called the SS stiffness (KSS). A high SS stiffness mimics stiff (i.e. less compliant) support surfaces and a low SS stiffness mimics more compliant support surfaces.
Figure 1: The Bilateral Ankle Perturbator (BAP) consists of two pedals, each driven by an electro-motor. The two pedals together form the support surface (SS) which can be controlled using a SS stiffness such that it mimics standing on foam. Rotating the SS with specific SS rotation amplitudes around the subjects ankle can evoke sensory reweighting of proprioceptive information. (B) The control scheme of each support surface of the Bilateral Ankle Perturbator (BAP) is shown in com-bination with the human. The BAP dynamics include a virtual inertia Iv, a virtual damping Bv and a virtual stiffness (i.e. support surface stiffness KSS). The human is represented by the dynamics of the rigid body (HIP) and the stabilizing mechanism. The stabilizing mechanism contains the vestibular (Wg), visual (Wv) and proprioceptive system (WP), the neural controller (NC) (including a time delay) and intrinsic dynamics due to the muscle properties.
Disturbance signal
A 20-second multisine disturbance was generated with frequencies in the range of 0.05-10 Hz containing 41 logarithmically distributed frequencies. Signal power was constant up to 3 Hz after which signal power decreased exponentially with frequency. The disturbance torque (Td) was divided into two and applied to both pedals simultaneously (Figure 2). Previous studies found that in normal stance, humans lean slightly forward and exert a total ankle torque of approximately 40 Nm [12]. Therefore, an additional torque of 20 Nm was added to the disturbance signal for each pedal of the SS, resulting in normal stance of the subjects when standing on the SS.
Process
Subjects were asked to stand on the BAP, without shoes and their arms crossed over their chest. Subjects wore a safety harness to prevent falling, which did not constrain movements or provided support in any way.
The experiment consisted of eleven trials, each containing 9 repetitions of the disturbance signal resulting in trials of three minutes (9 times 20 seconds). In the first nine trials, performed with eyes open, a combination of three levels of SS stiffness and three levels of SS rotation amplitude were applied.
The level of SS stiffness was chosen such that a high level (KSS,H = 700 Nm/rad) is comparable to the required stiffness to maintain an upright stance [13], a low level (KSS,L = 100 Nm/rad) is the minimum stiffness on which subjects are able to maintain balance (as determined in a pilot experiment) and a medium level (KSS,M = 300 Nm/rad) is set between the two extremes. Virtual damping and inertia were set to low values (i.e. B V of 24.2, 76.7 and 117.1 Nms/rad and virtual inertia IV of 0.2 kgm2).
Sensory reweighting of proprioceptive information was evoked by applying different levels of disturbance torques (Td), equal for all subjects. Three levels of disturbance torques were applied to generate three levels of rotations of the SS around the ankle axis (i.e. SS rotations (θSS)). These levels differed for each level of simulated SS stiffness, such that the resulting SS rotations had a peak-to-peak amplitude of approximately 0.03 (θSS,L), 0.07 (θSS,M) and 0.13 rad (θSS,H). These SS rotation amplitudes were comparable to previous sensory reweighting studies [1,3].
After the nine randomly performed trials with eyes open, the trial with medium SS rotation amplitude (θSS,M) combined with medium SS stiffness (KSS,M) and high SS stiffness (KSS,H) were performed with eyes closed resulting in a total of eleven trials. The combination of medium SS rotation amplitude (θSS,M) with low SS stiffness (K SS,L) was not included since pilot experiments indicated that most subjects were unable to perform this condition.
Data recording and processing
The applied torques to both support surfaces were summed to result in the disturbance torque (Td). The angles of both SS rotations (i.e. BAP motor angles) were measured and averaged to give the SS rotation amplitude (θSS). Total corrective ankle torque (Th) was obtained by the summation of the recorded torques (i.e. BAP motor torques) of both support surfaces. Two draw wire potentiometers (Celesto SP2-25, Celesto, Chatsworth, CA, United States) were attached to the right upper leg and the subjects’ trunk, measuring
the translations of the upper and lower body segments. All signals were recorded at a sample frequency of 1 kHz and processed in Matlab (The MathWorks, Natick, MA). The height of the Center of Mass (CoM) was calculated according to the equations of Winter et al. using the measured distance between the ground and ankle joint (lateral malleolus), the ankle and hip joint (greater trochanter), the hip and shoulder joint (acromion), and the subject’s height [14]. Body sway angle (θBS), i.e. angle of the Center of Mass (CoM) with respect to vertical was calculated based on potentiometer data and the height of the CoM.
Data analysis
The time series were segmented into data blocks of 20 seconds (i.e. the length of the disturbance signal) after which the first cycle was discarded because of transient effects, resulting in eight data blocks. Subsequently, for each subject and each trial, data were transformed to the frequency domain using the Fourier transform and averaged across the eight data blocks in the frequency domain. Cross spectral densities (CSD) between disturbance torque and body sway were calculated according to
(1)
in which Td(f) and θBS*(f) represent the Fourier transform of the disturbance torque and body sway. The asterisk indicates the complex conjugate. Similarly, CSDs were calculated between disturbance torque and corrective ankle torque and disturbance torque and SS rotation amplitude. CSDs were used to calculate the frequency response functions (FRFs) (equation 2 and 3). Two FRFs were estimated using closed-loop system identification methods [1,9]
(2)
(3)
The FRFs were only evaluated on the excited frequencies in the disturbance signal (f) up to 3 Hz. Equation (2) is the FRF of the torque sensitivity function (SSŜ
Th(ƒ)) to the disturbance, which describes the dynamic relation between the
proprioceptive disturbances (θSS) and the torque exerted by the ankles (Th) in terms of amplitude (magnitude) and timing (phase) as function of stimulus frequency [3]. The FRF of the sensitivity function contains dynamics of the rigid body and the stabilizing
Compliant support surfaces affect sensory reweighting Chapter 2
mechanism which comprises the visual (Wv), proprioceptive (Wp) and vestibular system (Wg), neural controller (NC) (including a time delay) and the intrinsic dynamics (Figure 1). The estimated FRF of the sensitivity function (SSŜ
Th(ƒ)) was normalized for the
gravitational stiffness, i.e. participants mass and the distance from the ankles to the CoM multiplied by the gravitational acceleration (mglCoM), which influences the FRF magnitude. A change in the FRF magnitude implies a relative change of responsiveness to the proprioceptive perturbations, i.e. sensory reweighting. The effects of SS stiffness and SS rotation amplitude on the FRF magnitude are most pronounced at the lower frequencies as the influence of sensory reweighting is most evident at low frequencies where system dynamics are dominated by sensory influences and are minimally affected by other factors such as inertia [1,3,10]. Therefore, the FRF magnitudes were averaged over the five lowest frequencies (0.05-0.25 Hz) resulting in a low frequency magnitude (LFM).
ĤIP(ƒ) (equation 3) is the estimated FRF of the rigid body dynamics, describing the
relation between the body sway (θBS) and the torque exerted by the ankles (T h) (Figure 1). To check whether this FRF was constant across all the trials for each subject and to validate the identification, the experimentally obtained FRF was compared to the theoretical transfer function of the rigid body dynamics (HIP(s)), which is representative
of an inverted pendulum and can be described with a moment of inertia (IIP), and a gravitational stiffness (mglCoM)
(4)
in which s denotes the Laplace operator, with s=i2πf, m is the mass of the subject, h the subjects height of the CoM relative to the ankle and g the gravitational acceleration [2]. Both the inertia and the CoM are derived according to Winter et al.[14].
Statistical Analysis
Linear mixed models were used statistically compare differences in LFM due to SS rotation amplitude and level of SS stiffness. SS rotation amplitude, SS stiffness and their interaction were included as covariates and set as fixed effects. The subject was included as a random effect to correct for differences in SS rotation amplitude due to variations in subjects corrective ankle torque and to take the measurement repetitions into account. For illustration purposes, regression lines were plotted between individual LFM and SS rotation amplitude for all levels of SS stiffness, together with the means and standard errors of both LFM and SS rotation amplitudes. To investigate the effect of closing the eyes, similar linear mixed models were used with eyes condition as additional covariate and fixed effect. For all tests, the significance level (α) was set at 0.05. All analyses were performed with SPSS version 22.0 (SPSS, Chicago, IL).
Figure 2: (Left) Time series of a typical subject for the trial with medium support surface (SS) stiff-ness and medium SS rotation amplitude, in which the disturbance torque (Td), SS rotation amplitude (θSS), corrective ankle torque (Th) and body sway angle (θBS) are shown. The mean of the time series is displayed in black and the standard errors in grey. For display purposes, signals were filtered with a phase preserving fourth order low pass digital Butterworth filter with cut-off frequency of 20 Hz. (Right) The associated power spectral densities are shown with the excited frequencies as dots.
Results
Time series
Figure 2 shows the time series as response to the disturbances of a typical subject in the medium KSS – medium θSS condition with eyes open. The standard error of the SS rotation amplitude is relatively low. The responses of body sway angle and corrective ankle torque are slightly more variable. The power spectral densities corresponding to the time series show that the excited frequencies are present in all signals.
Rigid body dynamics
Figure 3 shows the mean FRF and standard error of the FRF of the rigid body dynamics (ĤIP) of one typical subject, averaged over all trials with eyes open. Other subjects showed similar results. In addition, the theoretical FRF (HIP) according to the transfer function
described in the section ‘Data Analysis’ which is based on the subject’s anthropometries, is shown. The experimental FRF shows (up to 3 Hz) similar characteristics as the theoretical FRF although the magnitude is somewhat lower. Also, the standard error of the experimental FRF is low.
Figure 3: Frequency Response Function (FRF) of the rigid body dynamics (HIP) for one typical
subject is shown. The mean and standard error in the nine experimental trials with eyes open are show in black. The theoretical transfer function of the inverted pendulum, based on the body mass and height of the subjects Centre of Mass, is shown in grey.
Figure 4: Mean and standard error of the Frequency Response Functions (FRF) of the sensitivity function (SSS
TH), averaged over all subjects, normalized by gravitational stiffness. For each level
of support surface (SS) stiffness (low, medium and high KSS), the effect of SS rotation amplitude (low, medium and high θSS) on the FRF is shown. The vertical line indicates the lower frequency window over which statistical analysis is performed. The lower frequency magnitude (LFM) was calculated by averaging the magnitude of the five lowest frequency magnitudes from 0.05-0.25 Hz.
Sensitivity function
Figure 4 shows the estimated FRFs for each level of SS stiffness describing the sensitivity functions (SSŜ
Th(ƒ)), as an average across subjects with the SS rotation amplitude (in
trials with eyes open). Figure 5 shows, for all trials, the corresponding mean LFM and SS rotation amplitude averaged over all subjects with corresponding standard error of both SS rotation amplitude and LFM. In addition the regression line as function of the SS rotation amplitude is shown for each SS stiffness. The SS rotation amplitude is given as Root Mean Square (RMS) to eliminate the effect of outliers on the peak-to-peak amplitude. Due to the torque exerted by the human, the RMS of the SS rotation amplitude of all trials deviated from the three desired SS amplitude levels. Statistical analysis showed a significant interaction effect of SS rotation amplitude and SS stiffness on the LFM (p < 0.001) as shown in the slope of the regression lines. The slope of the regression line becomes smaller as the SS stiffness decreases, i.e. less sensory reweighting for a
Compliant support surfaces affect sensory reweighting Chapter 2
given change in SS amplitude. For each level of SS stiffness, linear mixed models showed a significant main effect of SS rotation amplitude (p < 0.001) on the LFM; by increasing SS rotation amplitude the LFM decreases.
Looking at the effect of closing the eyes, mean LFM and standard errors are shown in Figure 5. Linear mixed models showed a significant effect of closing the eyes (p<0.001) on the LFM. Closing the eyes resulted in a higher LFM. In addition, there was a significant effect of SS rotation amplitude (p=0.005) on the LFM similar to the previous trials. There was no significant effect of stiffness (p=0.088) and no interaction effect between SS stiffness and SS rotation amplitude (p=0.811).
Discussion
Sensory reweighting
As in previous studies, sensory reweighting is most pronounced at low frequencies where system dynamics are dominated by sensory influences and are minimally affected by other factors such as inertia [1,3,10]. In this study, we found a comparable effect of the SS rotation amplitude on the LFM; by increasing the SS rotation amplitude, the LFM decreased. This indicates that the stabilizing mechanism was down weighting the proprioceptive information accompanied by up weighting the vestibular and/or visual information as proprioceptive information was less reliable.
In addition, the trials with eyes closed showed a significant increase in LFM compared to eyes open conditions. This implies that closing the eyes (i.e. eliminating visual information) results in an up weighting of the proprioceptive information [1,3,10].
The effect of compliant support surfaces
Standing on compliant support surfaces, such as foam mats, is believed to disturb the proprioceptive information of the ankles by producing a time varying SS angle and making this information less reliable [7]. Due to the existence of the interaction effect of SS rotation amplitude and SS stiffness, the independent effect derived from SS stiffness is hard to interpret.
As for the interaction effect of SS stiffness and SS rotation amplitude, results showed that by decreasing SS stiffness, the slope of the regression line becomes smaller. This means that sensory down weighting of proprioceptive information as a response to increasing SS rotation amplitude is less when standing on more compliant support surfaces.
Figure 5: (A) Mean and standard errors of both lower frequency magnitude (LFM) of the sensitivity function (SSS
TH) and support surface (SS) rotation amplitudes, averaged over all subjects for each
level of SS stiffness (KSS). In addition, three regression lines, fitted on all individual data, are plotted for each level of SS stiffness. (B) Mean and standard errors of both LFM and SS rotation amplitudes, averaged over all subjects as function of SS rotation amplitude for the condition with medium (KSS Medium) and high (KSS High) SS stiffness combined with medium SS rotation amplitude with eyes closed and open.
Thus, sensory reweighting is relatively less when the proprioceptive information is already perturbed by a compliant SS independent of the effect of changes in SS rotation amplitude. The overall results of this study imply that when balance control is trained using foam mats, two different effects should be kept in mind. A foam mat with higher compliance (lower stiffness) leads to larger SS rotations and thus down weighting of proprioceptive information. However, the amount of down weighting for a given amount of SS rotation will also depend on the compliance of the foam mat itself.
Methodological considerations
The experimental FRF shows (up to 3 Hz) similar characteristics as the theoretical FRF although the magnitude is somewhat lower; humans behave more or less as an inverted pendulum. Also, the standard error of the experimental FRF of the rigid body dynamics is low, indicating that the rigid body dynamics do not change over the conditions within a subject. This implies that the changes in the FRF of the human (SSŜ
Th(ƒ)) are solely due to
